Towards Computational Linguistics in Minangkabau Language: Studies on Sentiment Analysis and Machine Translation

Towards Computational Linguistics in Minangkabau Language: Studies on Sentiment Analysis and Machine Translation Fajri Koto School of Computing and Information System The University of Melbourne ffajri@student.unimelb.edu.au Ikhwan Koto Faculty of IT Andalas University ikhwan220397@gmail.com Abstract Although some linguists (Rusmali et al., 1985; Crouch, 2009) have fairly attempted to de- fine the morphology and syntax of Minangk- abau, information processing in this language is still absent due to the scarcity of the an- notated resource. In this work, we release two Minangkabau corpora: sentiment analy- sis and machine translation that are harvested and constructed from Twitter and Wikipedia.1 We conduct the first computational linguistics in Minangkabau language employing classic machine learning and sequence-to-sequence models such as LSTM and Transformer. Our first experiments show that the classification performance over Minangkabau text signif- icantly drops when tested with the model trained in Indonesian. Whereas, in the ma- chine translation experiment, a simple word- to-word translation using a bilingual dic- tionary outperforms LSTM and Transformer model in terms of BLEU score. 1 Introduction Minangkabau (Baso Minang) is an Austronesian language with roughly 7m speakers in the world (Gordon, 2005). The language is spread under the umbrella of Minangkabau tribe – a matrilineal culture in the province of West Sumatra, Indone- sia. The first-language speakers of Minangkabau are scattered across Indonesian archipelago and Negeri Sembilan, Malaysia due to “merantau” (migration) culture of Minangkabau tribe (Drakard, 1999). 1Our data can be accessed at https://github.com/ fajri91/minangNLP Despite there being over 7m first-language speak- ers of Minangkabau,2 this language is rarely used in the formal sectors such as government and ed- ucation. This is because the notion to use Bahasa Indonesia as the unity language since the Indepen- dent day of Indonesia in 1945 has been a double- edged sword. Today, Bahasa Indonesia successfully connects all ethnicities across provinces in Indone- sia (Cohn et al., 2014), yet threatens the existent of some indigenous languages as the native speakers have been gradually decreasing (Novitasari et al., 2020). Cohn et al. (2014) predicted that Indonesia may shift into a monolingual society in the future. In this paper, we initiate the preservation and the first information processing of Minangk- abau language by constructing a Minangkabau– Indonesian parallel corpus, sourced from Twitter and Wikipedia. Unlike other indigenous languages such as Javanese and Sundanese that have been dis- cussed in machine speech chain (Novitasari et al., 2020; Wibawa et al., 2018), part-of-speech (Pratama et al., 2020), and translation system (Suryani et al., 2016), information processing in Minangkabau lan- guage is less studied. To the best of our knowledge, this is the first research on NLP in Minangkabau lan- guage, which we conduct in two different represen- tative NLP tasks: sentiment analysis (classification) and machine translation (generation). There are two underlying reasons why we limit our work in Minangkabau–Indonesian language pair. First, Minangkabau and Indonesian language 2In Indonesia, Minangkabau language is the fifth most spoken indigenous language after Javanese (75m), Sundanese (27m), Malay (20m), and Madurese (14m) (Riza, 2008). are generally intelligible with some …

Content truncated to 500 words